Source-Receptor Relationships for Ozone
and Fine Particulates in the Eastern
United States
Jhih-Shyang Shih, Alan J. Krupnick, Michelle
S. Bergin and Armistead G. Russell
May 2004 • Discussion Paper 04–25
Resources for the Future
1616 P Street, NW
Washington, D.C. 20036
Telephone: 202–328–5000
Fax: 202–939–3460
Internet: http://www.rff.org
© 2004 Resources for the Future. All rights reserved. No
portion of this paper may be reproduced without permission of
the authors.
Discussion papers are research materials circulated by their
authors for purposes of information and discussion. They have
not necessarily undergone formal peer review or editorial
treatment.
Source-Receptor Relationships for Ozone
and Fine Particulates in the Eastern United States
Jhih-Shyang Shih, Alan J. Krupnick, Michelle S. Bergin, and Armistead G. Russell
Abstract
A key question in developing effective mitigation strategies for ozone and particulate matter is
identifying which source regions contribute to concentrations in receptor regions. Using a direct
approach with a regional, multiscale three-dimensional model, we derive multiple sourcereceptor matrices (S-Rs) to show inter- and intrastate impacts of emissions on both ozone and
PM2.5 over the eastern United States. Our results show that local (in-state) emissions generally
account for about 23% of both local ozone concentrations and PM2.5 concentrations, while
neighboring states contribute much of the rest. The relative impact of each state on others varies
dramatically between episodes. In reducing fine particulate concentrations, we find that reducing
SO2 emissions can be 10 times as effective as reducing NOx emissions. SO2 reductions can lead
to some increase in nitrates, but this is relatively small. NOx reductions, however, lead to both
ozone reductions and some reduction in nitrate and sulfate particulate matter.
Keywords: source-receptor, ozone, particulate matter, sensitivity analysis, air quality
simulation, National Ambient Air Quality Standards
JEL Classification Numbers: Q2, Q25
Contents
Introduction............................................................................................................................. 1
Modeling Approach ................................................................................................................ 3
Model Description .............................................................................................................. 3
Application Description ...................................................................................................... 4
Model Performance............................................................................................................. 5
Results and Discussion............................................................................................................ 7
Sensitivity Results............................................................................................................... 7
Derivation of Multiple Pollutant S-R Matrices................................................................... 7
Ozone Sensitivity and Control Effects................................................................................ 9
PM2.5 Sensitivity and Control Effects ............................................................................... 11
Comparison of Long-Range Response of Ozone and PM2.5 Sensitivities ........................ 13
Issues of Nonlinearity and Uncertainty............................................................................. 14
Conclusions............................................................................................................................ 14
References.............................................................................................................................. 16
Source-Receptor Relationships for Ozone
and Fine Particulates in the Eastern United States
Jhih-Shyang Shih, Alan J. Krupnick, Michelle Bergin, and Armistead G. Russell ∗
Introduction
Controversy over the extent to which emissions in one state affect air quality in other states
has been a central feature of the debate over air pollution control policy since 1990. This
controversy culminated in a series of lawsuits brought by East Coast states against Midwestern
states, charging that the latter were making attainment of the National Ambient Air Quality
Standard for ozone much more difficult for the former. Additionally, given their influence in
creating fine particulate concentrations (PM2.5: particles with a diameter less than 2.5 microns),
the temporal and spatial contribution of sulfur dioxide (SO2) and oxides of nitrogen (NOx)
emissions and the effect of NOx emissions reductions on ozone must be understood if effective
air pollution policies are to be crafted.
Ozone and PM2.5 share common sources and formation routes in the atmosphere. Ozone, a
gas, is formed secondarily in the atmosphere by reactions between NOx and volatile organic
compounds (VOCs) in the presence of sunlight. Fine particulate matter is composed of many
different chemical species, including (but not limited to) ammonium sulfate, ammonium nitrate,
and organic and elemental carbon. Depending on the chemical composition, PM2.5 can be
directly emitted (e.g., from diesel vehicles, dust, and biomass burning) and/or formed
secondarily from reactions of sulfur dioxide, NOx, ammonia (NH3), and VOCs. Controls placed
∗ Jhih-Shyang Shih and Alan J. Krupnick are Fellow and Senior Fellow, respectively, at Resources for the Future. Michelle
Bergin and Armistead G. Russell are Ph.D. candidate and professor, respectively, at Georgia Institute of Technology. We thank
Yueh-Jiun Yang, Jim Wilkinson, Jim Boylan, Allen Basala, Sheau-Rong Lou, Dick Morgenstern and Kris Wernstedt for valuable
discussions. We are grateful to Lisa Crooks and Mike Batz for research assistance. This study is partly supported by a partnership
grant from the USEPA and NSF and through a USEPA STAR graduate fellowship. Note that superscripts refer to numbered
references on pages 16–17.
1
Resources for the Future
Shih, Krupnick, Bergin, and Russell
to reduce one form of particulate may affect concentrations of the other, though not necessarily
proportionally or even in the same direction. For example, reducing SO2 emissions will generally
lead to a reduction in sulfate particulate matter, though less than proportionally.1 Further, the
reduction in sulfate leads to more ammonia being available to form particulate ammonium
nitrate—i.e., there is a “rebound” in the amount of particulate formed. The two effects mitigate
the total PM reduction that might be first expected. Also investigated is the “bounce-back”
effect, where NOx reductions can lead to localized increases in ozone due to reduced scavenging
of ozone and ozone precursors. Typically, this happens within a few tens of kilometers of the
source region studied, with reductions found downwind.
Previous studies have explored the responses of ozone and of PM2.5 to various controls.2-5
Control strategies exist that lower both PM2.5 and ozone concurrently, or lower one or the other.
Thus, from an air quality management perspective, it is desirable to understand the relative
efficacy of NOx and SO2 emissions reductions in controlling these air pollutants.
To address these issues, we developed a new modeling approach to derive multiple sourcereceptor sensitivity matrices (S-Rs). 6 Independence, additivity, and linearity are assumed, and
have been demonstrated to be reasonably accurate.1 We compare the relative effectiveness of
regional (interstate) vs. local (intrastate) controls on NOx, and SO2 emissions as they affect ozone
and PM2.5 concentrations, the effectiveness of controlling NOx from elevated sources (i.e.,
electric utilities and large industrial plants), the relative effectiveness of reducing NOx vs. SO2
emissions to reduce PM2.5, and the sensitivity of these results to different weather patterns. We
study two very different meteorological episodes (in July and May 1995), each lasting
approximately 10 days. The 20 days allow us to investigate the impact of meteorological
variability on pollutant transport and local production. We also investigate the extent of the
“bounce-back” effect and the “rebound” effect, reducing part of the expected benefit from
controls. Finally, we investigate to what degree the efficacy of controls depends on how one
quantifies the response—e.g., looking at a spatial (area) weighting or considering exposure by
using a population-based approach.
2
Resources for the Future
Shih, Krupnick, Bergin, and Russell
Modeling Approach
Model Description
In this study, the Urban-to-Regional Multiscale (URM) One Atmosphere Model (URM1ATM)7-8 and the Regional Atmospheric Modeling System (RAMS) are used to account for the
processes significantly affecting ozone and PM2.5 concentrations in the atmosphere, including
atmospheric physics, gas and aerosol phase chemistry, cloud and precipitation processes, and wet
and dry deposition. RAMS is used to recreate the physics of an historical period of time,
providing details and spatial coverage unavailable from observations. URM-1ATM solves the
atmospheric diffusion equation (ADE) for the change in concentration, c, of species i with time,
∂ci
+ ∇ • (uci ) = ∇ • (K∇ci ) + f i + S i
∂t
(1)
where u is a velocity field, K is the diffusivity tensor, fi represents the production by chemical
reaction of species i, and Si represents sources and sinks of species i. As used here, a direct
sensitivity capability using the Direct Decoupled Method in Three Dimensions (DDM-3D)1-2 is
employed to calculate the local sensitivities of specified model outputs simultaneously with
concentrations. As shown in Equation 2, the sensitivity, Sij, of a model output, Ci (such as
pollutant concentration of species i) to specified model inputs or parameters, Pj (e.g., elevated
NOx emissions) is calculated as the ratio of the change in output Ci to an incremental change of
input or parameter Pj.
S ij =
∂Ci
∂Pj
(2)
This leads to an additional set of equations that are solved concurrently with the
concentrations, but the structure of those equations is very similar to the original (Equation 1),
and is solved efficiently. This sensitivity is a local derivative, so a linear assumption is then made
to extrapolate the result to a non-zero perturbation. This assumption has been well tested for the
3
Resources for the Future
Shih, Krupnick, Bergin, and Russell
variables of interest for this study.1 A more detailed description of the model is available
elsewhere.8-9
Application Description
We use a multiscale grid structure encompassing the eastern United States (Figure 1). The
finest grids are placed over major source regions such as the Ohio River Valley, where many
power plants and large industries are located, and over highly populated regions such as the East
Coast corridor. This approach allows evaluation of potential population exposure to pollutants
and captures high-population-related sources such as automobile exhaust, fast food restaurants,
and so forth. Here, the term “potential exposure” denotes that we do not take into account microenvironmental variations in concentrations or variations in individual activities. The vertical grid
has seven nonuniform layers, with the thinnest layer near the ground and the layers thickening
with an increase in altitude. This vertical scheme allows detailed treatment of ground-level
sources and best represents ground-level mixing and dry deposition processes, allowing for
diurnal changes in mixing depths, and capturing aloft multiday transport events. A comparison of
the effects of different grid scales and spatial allocation on model performance is presented
elsewhere.10
URM-1ATM is applied to two well-studied episodes occurring in July 9-19, 1995 and May
22-29, 1995. These base episodes were selected because high-quality and complete data was
available, was previously modeled using a different multiscale grid definition but with the same
simulation system,8 and because the data covered large meteorological variation with moderate
to high pollution formation. Meteorological information is developed using the Regional
Atmospheric Modeling System (RAMS)11 in a nonhydrostatic mode, including cloud and rain
microphysics. Prevailing winds in these episodes are toward the East Coast from the Midwest in
July, and toward the Southeast from the Mid-southern states in May. These are common wind
patterns for the summer and spring seasons, respectively.
4
Resources for the Future
Shih, Krupnick, Bergin, and Russell
Emissions were generated using the Emissions Modeling System (EMS-95).12 Because
1995 emissions are not relevant for future policy evaluation, 2010 day-specific emissions are
estimated under conditions that are anticipated with changes in population growth, vehicle
turnover, emissions control technologies, and anticipated emissions regulations. These future
scenarios are used to evaluate potential control strategies. Meteorological and initial and
boundary conditions are hourly and day-specific, and are held constant for base and future years.
Table 1a presents the estimated average day emissions for the 1995 basecases and the statespecific average day 2010 (future case) emissions. The emissions differences between 1995
basecases and the 2010 future case are due to implementation of the Tier 2 automotive emissions
regulations, the “NOx SIP Call,” acid deposition controls from the 1990 Clean Air Act
Amendments (CAAA), and other controls. Additional information on these model inputs is
described elsewhere.8-9
Model Performance
Both of the 1995 basecase episodes have been used to evaluate model performance against
ambient measurements. Excluding model ramp-up days, nine and six days of simulation are
available for evaluation in the July and May episodes, respectively. Data from the U.S.
Environmental Protection Agency’s Aerometric Information Retrieval System (AIRS)13 is used
to evaluate model performance for ozone prediction. EPA guidelines for urban scale ozone
modeling are +/- 15% for Normalized Mean Bias (NMB), and +/- 35% for Normalized Mean
Error (NME). This regional-scale application resulted in an average NMB for ozone of 3.2% for
the May episode and of 1.6% for the July episode; the NME was 16.8% for May and 21.4% for
July. All these results were well within the stated guidelines. These values are calculated from
measurements at sites coinciding with either 24km2 or 48km2 cells. Data was available from
nearly 500 sites for each episode.
The Interagency Monitoring of Protected Visual Environments (IMPROVE) network14
provides 24-hour averaged speciated aerosol data taken on Wednesday and Saturday of each
week in our episodes, and this data is used to evaluate model performance for aerosols. Three
5
Resources for the Future
Shih, Krupnick, Bergin, and Russell
days of data was available during the July episode (July 12, 15, and 19), with measurements
from 18 sites in total, 12 of which coincided with 24km2 or 48km2 cells. These sites resulted in
an average NMB for PM2.5 of -24.6% and an average NME of 31.67%. Two days of data was
available for the May episode (May 24 and 27), with measurements from 17 sites in total, 11 of
which coincided with 24km2 or 48km2 cells. These sites resulted in an average NMB for PM2.5 of
-9.4% and an average NME of 28.30%. There are no current guidelines to indicate acceptable
model performance for aerosols. A detailed description of ozone and speciated aerosol model
performance for this application is presented elsewhere.10
Figures 2a and 3a show sample results for 2010 ozone and PM2.5 concentrations for the
July episode. The predicted one-hour daily maximum ozone concentrations for the entire domain
for the July episode are between 84.5 ppb and 104 ppb, well below the peak levels simulated
using the 1995 emissions (which is 120 ppb), showing the impact of predicted future controls.
The ozone concentration reaches its peak on the fourth day of the episode. The predicted 24-hour
average daily maximum PM2.5 concentrations over the entire domain for the July episode are
mostly between 46.7µg/m3 and 70 µg/m3. The concentrations start exceeding 59 µg/m3 after the
third day, peak on July 14, and decrease again after July 15.
May’s lower temperature than July explains its lower ozone and PM2.5 concentrations. The
ozone concentration for the May episode is between 61.6 ppb and 89.5 ppb, about 20% lower, on
average, than the concentration during the July episode. The 24-hour average PM2.5
concentration during the May episode is between 25.4 µg/m3 and 65.5 µg/m3, about 22% lower
than the concentration during the July episode.
6
Resources for the Future
Shih, Krupnick, Bergin, and Russell
Results and Discussion
Sensitivity Results
After ensuring adequate model performance and developing future case simulations, the
2010 episodes are used to examine the sensitivity and control effects of ozone and aerosol
concentrations to reductions in specific emissions source types in different states. Here, S-R
coefficients are derived based on 30% emissions reductions. Nineteen states (or state
combinations) are evaluated as both sources and receptors. Results from reductions in two major
source types are presented: elevated NOx, such as from power plants and large industrial sites;
and total SO2 emissions, largely elevated emissions from coal-fired power plants. As examples,
Figures 2b and 3b show the “sensitivity” of ozone and PM2.5, respectively, resulting from
emissions reductions in Tennessee during the July 2010 episode. Figures 2b(1-3) show the
change in concentration of one-hour daily maximum ozone plumes at three different times of the
day when elevated NOx emissions are reduced by 30%. The impact on ozone concentrations in
other states is obvious (Figures 2b1 and 2b3). As is discussed later, some states, particularly
those near the Atlantic coast, are clearly receptors of many states’ emissions during the episodes
simulated (Figures 4-5).
Figure 3a shows the baseline results for PM2.5 concentrations during July 17, 2010, and
Figures 3b(1-3) show PM2.5 sensitivity to 30% SO2 emissions reductions from OH, KY, and IN.
The results clearly show the interstate impact of emissions reductions from these states.
Derivation of Multiple Pollutant S-R Matrices
The receptor states/regions of interest typically cover multiple simulation grid cells.
Therefore, to derive source-receptor matrices (S-Rs), we aggregated individual grid cell
sensitivity values to a single receptor site value. These sensitivity values represent the marginal
reduction of emissions from the source region to ozone or PM2.5 reduction at a receptor site. The
sensitivity used for aggregation is the change of pollutant concentration at the peak of the
specified time scale (one-hour or eight-hour daily maximum for ozone, and 24-hour average for
7
Resources for the Future
Shih, Krupnick, Bergin, and Russell
PM2.5). This source-receptor aggregation is performed on both a population-weighted and an
area-weighted basis. Population-weighted S-Rs are needed for estimating potential health
benefits from application of source controls, and also give a better proxy for health effects than
do area-weighted measures. From a regulatory perspective, the population-weighted S-Rs are
also more useful because they better apply to the urban areas (of course, under the Clean Air Act,
people living in rural areas are accorded the same level of protection as people in urban areas.)
The area-weighted S-Rs are useful to see the pure spatial and temporal effects of emissions on
concentrations.
Over the number of episode days, D, for the number of grid cells, i, covering receptor site
r, 1HDMsr is the area- or population-weighted one-hour daily maximum ozone source receptor
coefficient in change of ppb at receptor r per 1,000 tons/day of NOx emissions reduction from
source s.
I
D
∑∑ [∆O3 ]p
id
1HDM sr =
i =1 d =1
I
i
∆E s * D * ∑ pi
*1000
(3)
i =1
Here, ∆O3id is the daily maximum one-hour ozone concentration change in cell i and day d. The
daily average is calculated for each grid cell. For population weighting, this value is multiplied
by the cell population (pi) and divided by the total population, or pi is set to 1 for area weighting.
The value is then divided by the source precursor’s average daily emissions reduction in tons per
day, ( ∆E s ). This value is multiplied by 1,000 to convert it to ppb per 1,000 tons/day reduction.
This large factor is used for normalizing the sensitivities to avoid errors from very small
numbers.
While a per ton sensitivity comparison is useful for evaluating equivalent emissions
reductions between states, states have very large differences in what they emit (see Table 1b),
and therefore in how much they are able to reduce emissions. To account for this discrepancy,
we also define and present what we term “control effects” to quantify the pollution reduction in
receptor states that would be achieved by reducing emissions from the 19 states/regions
8
Resources for the Future
Shih, Krupnick, Bergin, and Russell
combined. Control effects are calculated by multiplying the sensitivity matrix by a vector of 30%
emissions reductions from each of the source states. We use population-weighted sensitivities in
these control effects calculations.
Ozone Sensitivity and Control Effects
There are four types of results combining the May and July episodes with area-weighted
and population-weighted sensitivities and control effects. We start with the July episode and
area-weighted sensitivities, and then compare these effects to those for May.
July Sensitivities. For the July episode and area-weighted S-Rs, states generally have only a
limited ability to reduce their own ozone concentrations by reducing elevated NOx emissions
(Table 2), although their own share of ozone reductions per unit NOx reduction is the single
largest for 53% of states (Figure 4). Results show that the average contribution of elevated NOx
control to local one-hour ozone concentrations is about 23%. The range of local control
contribution for elevated NOx is between 12% (MO) and 67% (WI). Note that emissions from
states outside of the 19 evaluated for sensitivities are not considered, and hence are not
contributing to this analysis. For this reason, a state such as Wisconsin may appear to be
contributing more to its own pollution than it actually is. The top five states contributing to
ozone concentrations in their own and other states (on a per unit emissions basis) are TN, KY,
VA, WV, and NC. The aggregate contributions to ozone per unit NOx reductions from these five
states to the rest of the domain are 12.4%, 8.7%, 8.5%, 7.3%, and 6.3%, respectively. In general,
downwind states (given our July conditions), such as MACTRI (MA, CT, and RI, combined),
MI, NY, MO, and NJ, contribute less to other states’ pollution.
We also calculate the sensitivities of eight-hour daily maximum ozone concentrations
consistent with the new U.S. ozone standard. The new standard is based on the fourth highest
eight-hour running average concentration of ozone in a year, while the historical standard is
based on the one-hour daily maximum. In general, the sensitivities of eight-hour daily maximum
9
Resources for the Future
Shih, Krupnick, Bergin, and Russell
ozone concentrations to NOx changes are slightly smaller than those for the one-hour daily
maximum ozone concentrations—2.5% smaller on average for elevated source NOx reductions.
To calculate health benefits obtained from emissions reductions, one would need to know
the change in population exposure. Since population is not uniformly distributed within
individual receptor state/regions, population-weighted sensitivities should be used in such
calculations instead of area-weighted sensitivities. For July, our population-weighted one-hour
NOx to ozone sensitivities are about 17.7% larger on average than area-weighted sensitivities,
with more than 74% of the local population-weighted sensitivities exceeding the corresponding
area-weighted sensitivities. Some of the differences can be quite striking. For example, for the
NJ receptor site, the one-hour ozone population-weighted sensitivity for NJ elevated NOx is
almost 40% bigger than that of its area-weighted counterpart (4.7 ppb vs. 3.5 ppb).
July Control Effects. Because the sensitivities were calculated based on a 30% reduction in
emissions, we estimate control effects for this same percentage change. This change, while
substantial, is lower than reductions proposed by the EPA Interstate Air Quality Rule, which
imply reductions in NOx and SO2 around 75% by 2015, on top of planned reductions already in
our 2010 baseline.
We find that the aggregate maximum one-hour ozone reduction ranges from 0.18 ppb (WI)
to 1.54 ppb (DE+MD) when elevated source emissions are reduced. These reductions may be
compared to baseline estimated ozone concentrations of 67.3 ppb (WI) and 77.2 ppb (DE+MD).
These reductions appear to be a small fraction of baseline ozone concentrations, in part because
emissions outside of the 19 states (other states and the domain boundaries) are not reduced,
contributing to elevated ozone levels.
July vs. May. Lower May temperatures and other differences in meteorological conditions
between May and July should result in lower predicted ozone concentrations in May, lower
ozone sensitivities to NOx in May, and different distributions of ozone contributions across states
(as seen in Figure 3). Indeed, the spatial and temporal average predicted one-hour daily
maximum ozone concentration in May is 53.9 ppb, while in July the average is 61.7 ppb. Also,
10
Resources for the Future
Shih, Krupnick, Bergin, and Russell
the average area-weighted one-hour daily maximum ozone sensitivity with respect to unit point
source NOx reductions for all sources and receptors for the May episode is 0.39 ppb, while it is
0.53 ppb for July, about 36% larger.
Less predictably, the local ozone sensitivity to elevated NOx reductions is higher for the
May episode (26% vs. 23% for July, on average). This implies that local control is more
effective in May compared to July. Because the winds come predominantly from the South in
May and from the Midwest in July, the top contributing states to other states’ ozone (as
measured by ozone sensitivities) tend to be from the South in May versus the Midwest in July
(Figure 3). It can also be seen that the variation among states’ contributions is larger in May.
PM2.5 Sensitivity and Control Effects
Fine particulate matter (considered here to be PM2.5) contains a mix of primary and
secondary components. In this paper, our discussion focuses on the sensitivity and control effects
of 24-hour averages of secondary PM2.5 only, starting with the July area-weighted sensitivities
with respect to sulfur dioxide and elevated source NOx emissions. The choice of a 24-hour
concentration averaging period is largely driven by the availability of measured data that was
available for model evaluation, and by the U.S. air quality standard, which is based on 24-hour
and one-year averages.
July Sensitivities. PM2.5 area-weighted sensitivities with respect to SO2 are far greater than those
with respect to NOx. For all sources and receptors during the July episode, the PM2.5 average
sensitivity with respect to total SO2 emissions is about 10 times the sensitivity with respect to
elevated point NOx. This is calculated from the average of the elemental ratio of two sourcereceptor matrices, namely the PM2.5 source-receptor matrix with respect to total SO2 emissions
divided by the PM2.5 source-receptor matrix with respect to elevated point NOx. As a specific
example, OH contributes 0.57 µg/m3 of PM2.5 to PA per 1,000 tons of SO2 emissions (on
average), but NOx emissions from OH elevated point sources contribute only 0.047 µg/m3 to PA
11
Resources for the Future
Shih, Krupnick, Bergin, and Russell
(per 1,000 tons). Thus, reducing a ton of total SO2 emissions in OH will have about 12 times the
impact on PM2.5 concentrations in PA as would reducing a ton of OH point NOx emissions.
As expected, we found that decreasing SO2 always decreases sulfates and increases
nitrates, with the former change being about 28 times the latter for the July episode. Our PM2.5
sensitivities represent the net effect of SO2 emissions control (including sulfate aerosol reduction
and slight nitrate aerosol increases). West et al.15 studied marginal PM2.5 under an equilibrium
model to estimate the conditions for nonlinear response to changes in sulfate concentrations.
Their study considered local effects (without pollutant transport). They found that reductions in
SO2 emissions may cause an increase in aerosol nitrate (the “bounce-back” effect) because nitric
acid converts to nitrate. Such conditions are found to be common and significant in winter—
perhaps three to four times as prevalent as in summer. As West et al.15 pointed out, the bounceback effect can be so large that PM2.5 concentrations can actually increase. In our study (as in
others),8 we found that reducing SO2 emissions reduces PM2.5 concentrations, and that the
bounce-back effect is small. Consideration of winter periods might find a larger formation of
nitrate aerosol.
July Control Effects. By reducing elevated NOx emissions 30% from each of the 19 evaluated
states/regions, the aggregate PM2.5 reduction in these states ranges from 0.006 µg/m3 (MO) to
0.071 µg/m3 (KY), and are derived from small reductions in particulate nitrate, sulfate, and
ammonium. These reductions are small largely because nitrate levels during the summer in this
region are small. SO2 sensitivities are much larger. By reducing total SO2 emissions 30% from
each of these states, the aggregate reduction ranges from 0.13 µg/m3 (WI) to 2.55µg/m3 (VA).
The reduction contributed by source states is not equally distributed, depending on the emissions
reduction and S-R coefficients between sources and receptor states. Virtually all of the reduction
is due to decreases in sulfate and ammonium, and there is usually a concurrent increase in nitrate.
These results may be compared to the annual PM2.5 standard of 15 µg/m3 and the daily standard
of 65 µg/m3 (the average over three consecutive years of the 98th percentile of the daily value of
PM2.5). Also, note that states such as WI and MO are on the upwind side of the region being
12
Resources for the Future
Shih, Krupnick, Bergin, and Russell
studied, so are affected by emissions not accounted for in this analysis. For the July episode, our
24-hour daily average PM2.5 concentrations range from 4.7 µg/m3 to 32.7 µg/m3.
July versus May. We find that predicted PM2.5 concentrations in May are low, averaging 12.0
µg/m3, while in July they average 16.5 µg/m3. Also, the average 24-hour area-weighted PM2.5
sensitivities with respect to SO2 reductions for all sources and receptors for the May episode are
about one-fifth the average for July: 0.032 µg/m3 versus 0.17 µg/m3. This is because, compared
to the July episode, the May episode had faster winds, lower temperatures, more rain, and lower
concentrations of biogenic VOCs and hydroxyl radicals.
On the other hand, the fraction of local contributions to PM2.5 reductions is higher for the
May episode (29.0% vs. 23%, on average). In addition, because the winds come predominantly
from the South in May and from the Midwest in July, the top contributing states to other states’
PM2.5 tend to be from the South in May versus the Midwest in July. As with ozone, the variation
among states’ contributions is larger in May (Figure 5).
Comparison of Long-Range Response of Ozone and PM2.5 Sensitivities
In this section, we examine how distance affects ozone and PM2.5 sensitivities. A
regression analysis was performed to explain ozone and PM2.5 sensitivities in terms of distance
from the source-state centroid to the receptor-state centroid. The dependent variable, r ij , is the
remote sensitivity normalized by the source-state local sensitivity.
r ij = β 0 + β 1 * d ij + ε
(4)
The right hand side, d ij , is the distance between the ith source-state centroid and the jth
receptor-state centroid. β 0 is the constant term of the regression equation. The slope ( β 1 ) of this
simple regression model can be interpreted as the decreasing sensitivity per unit of distance. ε is
the stochastic error term. On average for the July episode, the PM2.5 sensitivity to elevated source
NOx emissions reductions is significant, decreasing about 0.053% per km; the ozone sensitivity
decreases about 0.040 % per km. For the May episode, the PM2.5 sensitivity reduction is 0.041%
13
Resources for the Future
Shih, Krupnick, Bergin, and Russell
per km and the ozone sensitivity reduction is about 0.037 % per km. Results of this regression
are shown in Table 3.
Issues of Nonlinearity and Uncertainty
There are a number of important caveats that apply to the results of this study. First, we
assume that linearity holds within the 30% range of emissions changes and note that this
assumption was tested and found to be acceptable.1,6 Nevertheless, if the control measures were
to lead to emissions reductions much larger than 30%, the assumed linearity of response to
controls may no longer be satisfactory. In such a case, a second calculation at or near the reduced
emissions level may be necessary to account for the nonlinearity in pollutant response. Second,
we did not consider other sources of uncertainty and variability. For example, we only
considered two meteorological episodes (a total of 15 simulation days for use), and did not
extensively study the effect of meteorological input variation. While we have captured a high
and moderate pollutant episode, variation in wind direction, temperature, and other parameters
may significantly affect sensitivity results.
Conclusions
While the absolute effects of NOx and SO2 control on ozone and PM2.5 concentrations are
sensitive to meteorological conditions, reducing elevated NOx emissions from our 19-state study
area by 30% from 2010 levels will have, at most, a modest impact on ozone concentrations and
PM2.5 concentrations in most of those states. Note, however, that boundary emissions and
emissions from other states in the domain were not eliminated. Comparatively, reducing SO2
emissions can have significant effects on PM2.5 concentrations. While “rebound” and “bounceback” effects are observed, they are generally not significant enough to result in perverse effects
on ozone during these episodes. The rebound effect on PM2.5 of SO2 reductions is generally small
(in terms of nitrate increases).
14
Resources for the Future
Shih, Krupnick, Bergin, and Russell
Long-range transport of precursors is an important issue for both pollutants, although
perhaps less so than is commonly thought. Local contributions to air pollution problems account
for about 23% of total ozone concentrations and PM2.5 concentrations, and neighbouring states
contribute much of the rest. The ozone sensitivity results are comparable with others in the
literature based on simpler models.16-18 The PM sensitivity coefficients, however, are smaller.18
This result may be due, in part, to the episodes chosen for study.
Area and population-weighted sensitivity matrices tell very different stories. We found that
population-weighted sensitivities exceed area-weighted sensitivities by as much as six times in
some regions. When evaluating health damages, one should use population-weighted sensitivities
to account for human exposure.
15
Resources for the Future
Shih, Krupnick, Bergin, and Russell
References
1.
Odman, M.T., J.W. Boylan, J.G. Wilkinson, and A.G.Russell. Integrated Modeling for
Air Quality Assessment: The Southern Appalachian Mountains Initiative Project. J. Phys.
IV. 2002, 12 (PR10): 211-234
2.
Russell, A.G., K.F. McCue, and G.R. Cass. Mathematical Modeling of the Formation of
Nitrogen-Containing Pollutants—2. Evaluation of the Effects of Emission Controls.
Environ. Sci. Technol. 1998, 22, 1336-1347.
3.
Russell, A.G., K.F. McCue, and G.R. Cass. Mathematical Modeling of the Formation of
Nitrogen-Containing Air Pollutants—1. Evaluation of an Eulerian Photochemical Model.
Environ. Sci. Technol. 1998, 22, 263-271.
4.
Meng, Z., D. Dabdub, and J.H. Seinfeld. Chemical Coupling Between Atmospheric
Ozone and Particulate Matter. Science 1997, 277 (5322), 116-119.
5.
Stockwell, W.R., J.B. Milford, G.J. McRae, P. Middleton, and J.S. Chang. Nonlinear
Coupling in the NOx-SOx-Reactive Organic System. Atmos. Environ. 1988, 22, 24812490.
6.
Yang, Y.J., J.W. Wilkinson, and A.G. Russell. Fast, Direct Sensitivity Analysis of
Multidimensional Photochemical Models. Environ. Sci. Technol. 1997, 31, 2859-2868.
7.
Kumar, N., M.T. Odman, and A.G. Russell. Multiscale Air Quality Modeling:
Application to Southern California. J. Geophys. Res. 1994, 99, 5385-5397.
8.
Boylan, J.W., M.T. Odman, J.G. Wilkinson, A.G. Russell, K. Doty, W. Norris, and R.
McNider. Development of a Comprehensive, Multiscale “One Atmosphere” Modeling
System: Application to the Southern Appalachian Mountains. Atmos. Environ. 2002, 36,
3721-3734.
9.
Bergin, M.S, J-S Shih, J.W. Boylan, J.G. Wilkinson, A.J. Krupnick, and A.G. Russell.
Inter- and Intra-State Impacts of NOx and SO2 Emissions on Ozone and Fine Particulate
Matter. To be submitted to Environ. Sci. Technol (2004).
16
Resources for the Future
10.
Shih, Krupnick, Bergin, and Russell
Bergin, M.S, J.W. Boylan, J.G. Wilkinson, J-S Shih, A.J. Krupnick, and A.G. Russell.
Effects of Multiscale Grid Resolution and Spatial Distribution on 3D Model Performance
for Ozone and Aerosols. To be submitted to Environ. Sci. Technol (2004).
11.
Pielke, R.A., W.R. Cotton, R.L. Walko, C.J. Tremback, W.A. Lyons, L.D. Grasso, M.E.
Nicholls, M.D. Moran, D.A. Wesley, T.J .Lee, and J. H. Copeland, A Comprehensive
Meteorological Modeling System - RAMS. Meteor. Atmos. Phys. 1992, 49, 69-91.
12.
Wilkinson, J.G., C.F. Loomis, D.E. McNally, R.A. Emigh, and T.W. Tesche. Technical
Formulation Document: SARMAP/LMOS Emissions Modeling System (EMS-95). AG90/TS26 & AG-90/TS27. Alpine Geophysics, Pittsburgh, PA (1994).
13.
USEPA EPA AIRS Data. U.S. Environmental Protection Agency, Office of Air Quality
Planning & Standards, Information Transfer & Program Integration Division,
Information Transfer Group. www.epa.gov/airsdata (2001).
14.
NPS. National Park Service, Air Quality Research Division, Fort Collins. Anonymous ftp
at ftp://alta_vista.cira.colostate.edu in /data/improve (2000).
15.
West, J.J., A.S. Ansari, and N.P. Spyros. Marginal PM2.5: Nonlinear Aerosol Mass
Response to Sulfate Reductions in the Eastern United States. J. Air & Waste Manage.
Assoc. 1999, 49, 1415-1424.
16.
Rao, S.T.; Mount, T.D.; Dorris, G. Least Cost Control Strategies to Reduce Ozone in the
Northeastern Urban Corridor. The NY State Department of Environmental Conservation,
Division of Air Resources, Albany, NY (1999).
17.
Krupnick, Alan J., Virginia D. McConnell, David H. Austin, Matthew Cannon, Terrell
Stoessell, and Brian Morton. The Chesapeake Bay and the Control of NOx Emissions: A
Policy Analysis, RFF discussion paper 98-46, August, Washington, DC. (1998).
18.
Krupnick, A., V. McConnell, M. Cannon, T. Stoessell, and M. Batz. Cost-Effective NOx
Control in the Eastern United States, RFF discussion paper 00-18, August, Washington,
DC. (2000).
17
Resources for the Future
Shih, Krupnick, Bergin, and Russell
Table 1(a). Basecase domain-wide 1995 average day emissions (tons/day)
Emission
NOx
SO2
Organic
Gases
Source Type
Area
Biogenic
Power plants
Non-road Mobile
On-road Mobile
Elevated Point
Power plants
Other
Biogenic
Anthropogenic
July 11-19
May 24-29
3,008
5,284
22,248
7,319
12,509
8,110
34,140
20,094
182,381
46,014
5,573
3,024
17,063
5,932
12,855
7,425
26,364
20,259
76,315
43,467
18
Resources for the Future
Shih, Krupnick, Bergin, and Russell
Table 1(b). State-specific 2010 average day emissions (tons/day)
July 11-19
NOx
State Elevated
SO2
Ground
Total
May 24-29
Organic Gases
Biogenic
NOx
Anthroa
Elevated
level
AL
DC
GA
IL
IN
KY
MA
+
CT
+RI
MD
+
DE
MI
MO
NC
NJ
NY
OH
PA
SC
TN
VA
WI
WV
a
SO2
Ground
Organic Gases
Total
Biogenic
Anthroa
level
343
1
340
667
516
292
644
48
649
1,489
1,033
653
1,487
26
2,043
2,268
2,514
1,762
11,457
7
9,412
2,870
2,239
5,491
1,051
30
1,162
1,988
1,394
785
338
1
335
661
505
288
710
50
969
1,342
887
591
1,493
25
1,977
2,442
2,495
1,761
6,275
3
6,017
563
555
2,097
1,152
30
1,494
1,897
1,319
767
120
609
841
3,083
1,090
116
526
875
1,798
934
172
544
234
317
90
249
497
549
246
216
186
186
331
406
902
682
606
461
795
1,033
738
423
891
748
601
273
852
1,343
1,066
1,972
748
1,559
2,732
1,388
942
1,098
1,376
1,159
1,550
1,285
3,541
12,545
6,961
901
2,130
2,537
4,852
4,420
7,124
7,014
5,283
3,916
604
1,409
1,082
1,461
844
1,558
1,542
1,231
1,050
1,619
1,250
891
415
164
546
231
311
90
231
483
533
242
212
180
181
324
430
988
532
739
578
857
989
870
438
859
737
548
285
928
1,399
1,033
1,936
951
2,164
2,650
2,078
931
1,092
1,401
1,224
1,543
497
842
2,718
3,475
359
841
737
1,465
2,690
3,131
2,425
1,077
1,341
605
1,454
1,073
1,590
862
1,549
1,449
1,179
1,072
1,587
1,254
873
401
Anthro: Anthropogenic VOC
19
Resources for the Future
Shih, Krupnick, Bergin, and Russell
Table 2. Summary of state contributions (%) to local air quality
per unit emissions reduction (July episode, area-weighted)
1-hr Ozone
8-hr Ozone
Reduced
Reduced
Reduced
Reduced
Elevated NOx
Elevated NOx
Elevated NOx
Total SO2
AL
36
34
35
36
DE+MD
21
20
12
14
GA
24
23
24
27
IL
13
10
11
19
IN
19
17
18
22
KY
25
25
22
23
MA+CT+RI
16
20
19
21
MI
12
13
16
15
MO
12
14
21
12
NC
22
21
19
22
NJ
21
21
10
15
NY
14
13
16
20
OH
15
15
6
15
PA
16
15
17
21
SC
32
32
33
42
TN
41
39
38
35
VA
24
24
21
16
WI
67
67
69
40
WV
17
17
10
25
Average
23
23
22
23
Receptor
State
Episode average 24-hr PM2.5
Local contribution to reductions in one-hour daily maximum ozone and eight-hour daily maximum ozone from unit
reductions in elevated NOx and reductions in 24-hour PM2.5 concentrations from unit reductions in elevated NOx and
total SO2 emissions. For example, if each state reduced elevated NOx emissions by one ton, then 36% of the
resulting ozone reductions in Alabama would be attributed to reductions of NOx within Alabama.
20
Resources for the Future
Shih, Krupnick, Bergin, and Russell
Table 3. Results of long-range response of ozone and PM2.5 sensitivities with respect to
elevated point NOx reduction for July and May episodes.
Regression
β0
β1
Adj R-squared
July ozone
July PM2.5
May ozone
May PM2.5
0.72
0.95
0.63
0.64
(13.95)
(10.63)
(17.86)
(21.13)
-0.00040
-0.00053
-0.00037
-0.00041
(-8.56)
(-6.56)
(-11.67)
(-14.92)
0.167
0.105
0.273
0.381
Numbers in parentheses are t-statistics. All slope coefficients are significant at the 95% level or better.
21
Resources for the Future
Shih, Krupnick, Bergin, and Russell
Figure 1. Multiscale grid used to model changes in ozone and particulate species from changes
in NOx and SO2. The finest resolution has horizontal grids of 24km per side, and the other cells
are 48km, 96km, and 192km per side. The shaded areas represent high population densities
(urban areas). Fine-scale cells are placed over areas of high industrial or population densities.
22
Resources for the Future
Shih, Krupnick, Bergin, and Russell
Figure 2. Ozone and its sensitivity to a reduction of elevated NOx emissions. (a) Concentration
of ozone simulated for July 14, 2010 at 1:00 p.m. (b) Change in ozone concentration at (b1)
10:00 a.m., (b2) 1:00 p.m., and (b3) 4:00 p.m. caused by a 30% reduction in elevated NOx
emissions from Tennessee.
23
Resources for the Future
Shih, Krupnick, Bergin, and Russell
Figure 3. PM2.5 24-hour average concentration and sensitivity to total SO2. (a) 24-hour average
concentration of PM2.5 simulated for July 17, 2010; (b) 24-hour average change in PM2.5
concentration caused by a 30% reduction in total SO2 emissions from (b1) Ohio, (b2) Kentucky,
and (b3) Indiana.
24
Resources for the Future
Shih, Krupnick, Bergin, and Russell
Figure 4. Relative area-weighted one-hour maximum average ozone sensitivities with respect to
a unit reduction in elevated NOx (ppb per 1,000 tons/day) for both the May and July episodes.
For each state, the top six contributing state/regions are shown, with the rest of the evaluated
states combined into an “other” category. For each receptor, the left column represents the results
for May and the right the results for July. The height of an individual component of a column
represents the percentage of ozone change caused by a unit NOx reduction. The color of the
component indicates the source (state/region) of that emission. For example, for the state of
Maryland during the July episode, the column indicates that the top six contributors to Maryland
ozone concentrations are Maryland, West Virginia, Virginia, Pennsylvania, Ohio, and North
Carolina. During the May episode, however, Maryland’s own contribution is lower, and
Tennessee and Kentucky become major contributors.
25
Resources for the Future
Shih, Krupnick, Bergin, and Russell
Figure 5. Relative area-weighted 24-hour average PM2.5 sensitivities with respect to a unit
reduction in total SO2 emissions (µg/m3 per 1,000 tons/day) for the May and July episodes. The
top six contributing state/regions and “all other states” are presented.
26